Anmerkungen:
Nach Informationen von SSRN wurde die ursprüngliche Fassung des Dokuments September 1, 2011 erstellt
Beschreibung:
This article demonstrates the use of two approaches to parallelizing a Garch(1,1) calibration algorithm. The base serial implementation is a genetic algorithm that uses maximum likelihood in the fitness function. This is written in generic C. The first type of parallelization involves calculating all the fitness functions of the population in parallel. The second type of parallelization involves parallelizing the fitness function itself. The parallel algorithms are developed in nVIDIA CUDA and run on a GPU architecture